Olympiad Algebra II: Polynomials, Recurrences & Radical Equations

Study Sheet

Olympiad Algebra II: Polynomials, Recurrences & Radical Equations

Vieta in reverse, Newton in miniature, characteristic roots, extraneous solutions

Algebra at Proof Level

Tip
Building polynomials from symmetric data

Given r+s+tr + s + t, rs+st+trrs + st + tr, rstrst, the monic cubic with those roots is x3−e1x2+e2x−e3x^3 - e_1x^2 + e_2x - e_3. Power sums convert through Newton: x2+y2=(x+y)2−2xyx^2 + y^2 = (x + y)^2 - 2xy, x3+y3=(x+y)3−3xy(x+y)x^3 + y^3 = (x+y)^3 - 3xy(x+y).

Sums 6,11,66, 11, 6 give (x−1)(x−2)(x−3)(x-1)(x-2)(x-3), whose value at 44 is 66; x+y=4x + y = 4, x3+y3=28x^3 + y^3 = 28 forces xy=3xy = 3 and x2+y2=10x^2 + y^2 = 10. Every "find this symmetric expression" problem is a two-step: reach e1,e2,(e3)e_1, e_2, (e_3), then rebuild. If the problem wants the actual roots, factor the rebuilt polynomial by testing small integers (rational root theorem).

Tip
Linear recurrences and their characteristic roots

an+2=pan+1+qana_{n+2} = pa_{n+1} + qa_n has general term Aαn+BβnA\alpha^n + B\beta^n where α,β\alpha, \beta solve r2=pr+qr^2 = pr + q (with (A+Bn)αn(A + Bn)\alpha^n for a double root).

an+2=an+1+2ana_{n+2} = a_{n+1} + 2a_n has roots 2,−12, -1: an=2n−(−1)n3a_n = \tfrac{2^n - (-1)^n}{3}, so a10=341a_{10} = 341. The method IS diagonalization of the companion matrix (Linear Algebra course), which is why it also handles kk-term recurrences and why a root of modulus >1> 1 dominates growth. Nonhomogeneous recurrences add a particular solution, exactly as for differential equations.

Tip
Radical and rational equations: the ghost problem
xy−4−224−4−22y = √(x+5)y = −√(x+5)y = x − 1ghost
Squaring solves √(x+5) = ±(x − 1) at once: the ghost x = −1 lies on the dashed branch.

Squaring, clearing denominators, and cross-multiplying can INTRODUCE solutions. Solve, then verify every candidate in the original equation.

x+5=x−1\sqrt{x + 5} = x - 1 squares to (x−4)(x+1)=0(x - 4)(x + 1) = 0; x=−1x = -1 is a ghost (the left side is 22, the right −2-2). The clean discipline: record the domain and sign constraints BEFORE squaring (x−1≥0x - 1 \ge 0 here), then squaring is reversible on that domain. For nested radicals, isolate one radical per squaring; for a±b\sqrt{a} \pm \sqrt{b} forms, multiply by the conjugate.

Side note
Substitutions that linearize

x+1x=tx + \tfrac1x = t collapses palindromic polynomials; u=x2u = x^2 halves the degree of even polynomials; x=y−a3x = y - \tfrac{a}{3} removes the x2x^2 term of a cubic (Cardano's first step); x=1−t1+tx = \tfrac{1 - t}{1 + t} and x=tan⁡θx = \tan\theta tame 1±x1∓x\tfrac{1 \pm x}{1 \mp x} and 1+x2\sqrt{1 + x^2}. Recognizing WHICH substitution the equation is asking for is the algebra half of the USAMO.

Try it
Try it: a palindromic quartic

Solve x4−5x3+6x2−5x+1=0x^4 - 5x^3 + 6x^2 - 5x + 1 = 0.

Show the answer

Divide by x2x^2: (x2+1x2)−5(x+1x)+6=0\left(x^2 + \tfrac{1}{x^2}\right) - 5\left(x + \tfrac1x\right) + 6 = 0. With t=x+1xt = x + \tfrac1x, x2+1x2=t2−2x^2 + \tfrac1{x^2} = t^2 - 2: t2−5t+4=0t^2 - 5t + 4 = 0, t=1t = 1 or 44. t=4t = 4 gives x=2±3x = 2 \pm \sqrt3; t=1t = 1 gives complex roots. Real solutions: 2±32 \pm \sqrt3.

Proofs & Why It Matters

Tip
Proof: the characteristic-root formula for two-term recurrences

If α≠β\alpha \ne \beta solve r2=pr+qr^2 = pr + q, every solution of an+2=pan+1+qana_{n+2} = pa_{n+1} + qa_n is Aαn+BβnA\alpha^n + B\beta^n.

Both αn\alpha^n and βn\beta^n satisfy the recurrence (substitute and factor out rnr^n). Solutions form a vector space closed under linear combinations, and a solution is determined by (a0,a1)(a_0, a_1) — two numbers — so the space is 2-dimensional. αn\alpha^n and βn\beta^n are independent (their initial pairs (1,α)(1, \alpha), (1,β)(1, \beta) are independent for α≠β\alpha \ne \beta), hence a basis. ■\blacksquare For a double root, nαnn\alpha^n is the second basis element — verify by substitution, exactly as for xerxxe^{rx} in differential equations.

Tip
Proof: the rational root theorem

If pq\tfrac pq in lowest terms is a root of anxn+⋯+a0a_nx^n + \cdots + a_0 with integer coefficients, then p∣a0p \mid a_0 and q∣anq \mid a_n.

Multiply ∑ak(p/q)k=0\sum a_k(p/q)^k = 0 by qnq^n: ∑akpkqn−k=0\sum a_kp^kq^{n-k} = 0. Every term except a0qna_0q^n is divisible by pp, so p∣a0qnp \mid a_0q^n, and gcd⁡(p,q)=1\gcd(p, q) = 1 forces p∣a0p \mid a_0. Symmetrically every term except anpna_np^n is divisible by qq, so q∣anq \mid a_n. ■\blacksquare Significance: it makes "factor by finding an integer root" a finite search, and its special case "a monic integer polynomial has no non-integer rational roots" is why 2\sqrt2 is irrational.

Going Deeper: Worked Problems

Example
Worked: a system solved by Vieta in reverse

Find all real triples with x+y+z=6x + y + z = 6, x2+y2+z2=14x^2 + y^2 + z^2 = 14, x3+y3+z3=36x^3 + y^3 + z^3 = 36.

Reminder — Vieta's formulas:  r1+r2=−ba,r1r2=car_1+r_2=-\frac{b}{a},\qquad r_1 r_2=\frac{c}{a}

Step 1 — e1=6e_1 = 6; e2=e12−p22=36−142=11e_2 = \tfrac{e_1^2 - p_2}{2} = \tfrac{36 - 14}{2} = 11; Newton: p3=e1p2−e2p1+3e3p_3 = e_1p_2 - e_2p_1 + 3e_3 gives 36=84−66+3e336 = 84 - 66 + 3e_3, so e3=6e_3 = 6.

Step 2 — the triple consists of the roots of t3−6t2+11t−6=(t−1)(t−2)(t−3)t^3 - 6t^2 + 11t - 6 = (t-1)(t-2)(t-3).

Step 3 — all permutations of (1,2,3)(1, 2, 3): six solutions. Symmetric systems never require solving for one variable at a time; rebuild the polynomial and factor.

Example
Worked: a recurrence with a nonhomogeneous term

Solve an+1=2an+3a_{n+1} = 2a_n + 3 with a0=1a_0 = 1, and find a10a_{10}.

Step 1 — the fixed point: L=2L+3L = 2L + 3 gives L=−3L = -3.

Step 2 — shift: bn=an+3b_n = a_n + 3 satisfies bn+1=2bnb_{n+1} = 2b_n, so bn=2nb0=4⋅2nb_n = 2^nb_0 = 4\cdot2^n.

Step 3 — an=4⋅2n−3=2n+2−3a_n = 4\cdot2^n - 3 = 2^{n+2} - 3; a10=4096−3=4093a_{10} = 4096 - 3 = 4093.

Step 4 — check: a1=2+3=5=8−3a_1 = 2 + 3 = 5 = 8 - 3 ✓. Subtracting the fixed point is the recurrence version of "particular solution plus homogeneous," and it works for any an+1=kan+ca_{n+1} = ka_n + c with k≠1k \ne 1.

Example
Worked: ⌊(2+3)n⌋\lfloor(2+\sqrt3)^n\rfloor is odd

Prove that ⌊(2+3)n⌋\left\lfloor(2+\sqrt3)^n\right\rfloor is odd for every positive integer nn. (folklore)

Step 1 — pair the number with its conjugate: sn=(2+3)n+(2−3)ns_n=(2+\sqrt3)^n+(2-\sqrt3)^n. Both 2±32\pm\sqrt3 are roots of r2=4r−1r^2=4r-1, so sns_n satisfies sn+1=4sn−sn−1s_{n+1}=4s_n-s_{n-1} with s0=2s_0=2, s1=4s_1=4.

Step 2 — hence every sns_n is an even integer: 2,4,14,52,194,…2,4,14,52,194,\dots (even starting values and an integer recurrence preserve evenness).

Step 3 — 0<2−3<10<2-\sqrt3<1, so 0<(2−3)n<10<(2-\sqrt3)^n<1 for n≥1n\ge1. Therefore (2+3)n=sn−(2−3)n(2+\sqrt3)^n=s_n-(2-\sqrt3)^n lies strictly between sn−1s_n-1 and sns_n, and its floor is sn−1s_n-1, which is odd. For example (2+3)2=7+43≈13.93(2+\sqrt3)^2=7+4\sqrt3\approx13.93 and s2=14s_2=14.

How to find it. Powers of a quadratic irrational are almost integers when the conjugate is small. The sum with the conjugate is an integer sequence satisfying the characteristic-root recurrence, and its parity is read off the recurrence.

Side note
Where it goes wrong: the double root

Solve an+2=4an+1−4ana_{n+2}=4a_{n+1}-4a_n, a0=1a_0=1, a1=4a_1=4. Tempting: "the characteristic equation r2−4r+4=0r^2-4r+4=0 has root 22, so an=A⋅2n+B⋅2na_n=A\cdot2^n+B\cdot2^n." Then a0=A+B=1a_0=A+B=1 and a1=2(A+B)=4a_1=2(A+B)=4 — contradictory.

A⋅2n+B⋅2nA\cdot2^n+B\cdot2^n is just (A+B)2n(A+B)2^n: one free constant, not two, so it cannot match two initial values. For a double root α\alpha the second basis solution is nαnn\alpha^n (substitute: (n+2)αn+2−4(n+1)αn+1+4nαn=αn(n(α−2)2+2α(α−2))=0(n+2)\alpha^{n+2}-4(n+1)\alpha^{n+1}+4n\alpha^n=\alpha^n\left(n(\alpha-2)^2+2\alpha(\alpha-2)\right)=0 at α=2\alpha=2). With an=(A+Bn)2na_n=(A+Bn)2^n: A=1A=1 and 2(1+B)=42(1+B)=4, so B=1B=1 and an=(n+1)2na_n=(n+1)2^n — indeed a2=4⋅4−4=12=3⋅4a_2=4\cdot4-4=12=3\cdot4.

Try it
Try it: an almost-integer

Show that (1+2)n+(1−2)n(1+\sqrt2)^n+(1-\sqrt2)^n is an even integer for every n≥0n\ge0, and use it to find ⌊(1+2)6⌋\left\lfloor(1+\sqrt2)^6\right\rfloor.

Show the answer

1±21\pm\sqrt2 are the roots of r2=2r+1r^2=2r+1, so sn=(1+2)n+(1−2)ns_n=(1+\sqrt2)^n+(1-\sqrt2)^n satisfies sn+1=2sn+sn−1s_{n+1}=2s_n+s_{n-1} with s0=2s_0=2, s1=2s_1=2: the terms are 2,2,6,14,34,82,1982,2,6,14,34,82,198, all even. For n=6n=6, (1−2)6=(2−1)6(1-\sqrt2)^6=(\sqrt2-1)^6 lies in (0,1)(0,1), so (1+2)6=198−(2−1)6(1+\sqrt2)^6=198-(\sqrt2-1)^6 is strictly between 197197 and 198198, and its floor is 197197. (Directly: (1+2)6=99+702≈197.99(1+\sqrt2)^6=99+70\sqrt2\approx197.99.)